A Consumer-Oriented Car Style Evaluation System Based on Fuzzy Mathematics and Neural Network

As an important link in product development, car style evaluation could ensure the quality of car style design, making the design more efficient, laying the foundation for production planners, production managers, and investment decision-makers in automobile manufacturing. The consumer-centered evaluation should accurately reflect the psychological cognition and subjective feelings of consumers. However, the current studies have not provided a unified evaluation standard, nor fully utilized the massive data on the evaluations made by consumers. Considering in advantages of fuzzy mathematics and neural network in processing massive data on consumer evaluations, this paper designs a consumer-oriented car style evaluation system based on these two techniques. Firstly, a scientific evaluation index system was designed for consumer-oriented car style evaluation, the index scores were classified into different levels, and a judgment matrix was constructed for indices on each layer and subject to consistency check. Next, absolute weights were assigned to alternatives, and the corresponding fuzzy membership functions were determined, producing a fuzzy comprehensive evaluation (FCE) model based on analytic hierarchy process (AHP) (AHP-FCE model) for car style evaluation. Furthermore, car styles were categorized by appearance structure, and the car style samples were parametrized for evaluation. Finally, particle swarm optimization (PSO) was improved, and then combined with backpropagation neural network (BPNN) into a classification model for consumer-oriented car style evaluation. The proposed consumer-oriented car style evaluation model was proved effective and superior through experiments. The results offer a reference for the application of the model in other evaluation scenarios